If this is meant to be a seamless experience starting with the truck driving in and transitioning into the 3D model, it’s going to be challenging. For one, I’d be surprised if 6 cameras would give a decent photogrammetry result. Even 36 cameras might require a good amount of cleanup IMO, but someone with more photogrammetry experience may have a better answer.
Is there an option to either keep it pure shoot or pure CG? As a shoot, it would require a motorized camera rig or bullet time rig to capture an arc of the car. It wouldn’t be 3D, and maybe not full 360, but it would be visually consistent. Obviously, it adds cost to the shoot. For full CG, access to the CAD models would probably be necessary, but again you’d have consistency between the car driving in and the 3D model, as they would be the same.
There is cool AI tech that reconstructs a full 3D model from a single photograph, but it may not be not accurate enough for an automotive client. Nvidia has a plug-in for Omniverse called GANverse 3D. This demo actually is using a car base model and single photographs to re-model it as far as I can tell. I’m not sure of the export process if any:
https://youtu.be/gz5E9wszZSIIf Nathan Shipley is here, he’s played with some similar underlying tech and may know of some other AI solutions for photo based model construction.